The 2 main objectives for this work are the development of a robust HTS assay for human GALK activity and the use of this established assay to identify small molecules that inhibited human GALK activity in vitro. Being the first enzyme of the Leloir pathway of galactose metabolism, the level of GALK activity dictates the amount of galactose entering into the galactose metabolic pathway, making GALK a significant biological target for overall galactose metabolism. Hence, discovery of inhibitors for this important biological target will not only provide an opportunity for novel therapy for CG but also advance our understanding of the role played by galactose catabolism in cellular metabolism.
Overexpression and purification of human GALK
Initial estimation showed that a minimum of 5 mg of purified GALK would be required for the development of the HTS assay, as well as the subsequent screening of 50,000 compounds in duplicate. Because overexpression and purification of active recombinant galactokinases had been successfully performed via affinity-chromatography using both E. coli
and yeast as hosts,30,31
we decided to adopt one of the published protocols to prepare sufficient human GALK enzyme for our screening needs. As shown in , we achieved a typical yield of 0.1 mg of purified GALK protein from 1 liter of bacterial culture. Galactose-dependent ATPase activity of the purified enzyme was verified using the standard pyruvate kinase/lactate dehydrogenase-coupled assay.25
We confirmed the presence of purified recombinant His6-tagged GALK protein using anti-His6-tag antibody (data not shown).
FIG. 2 Purification of human galactokinase (GALK). Different amounts of Ni2+-NTA affinity-purified GALK protein were loaded on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) as follows: M, molecular weight markers; 1, unloaded; (more ...)
Using the standard pyruvate kinase/lactate dehydrogenase-coupled assay, we established that the KM
for ATP was 42 μM and KM
for galactose was 210 μM in the purified human GALK. Timson and Reece,31
using the same assay, established that the KM
for ATP was 34 μM, and the KM
for galactose was 970 μM.31
The differences in the KM
values for galactose remain as yet unexplained but could be due to differences in ionic strength, temperature, and the percentage of the α-anomeric form (vs. β-anomeric form) of the D-galactose used.
Development of HTS assay for purified GALK
Despite the relatively low yield of GALK from our bacterial expression system, we harvested sufficient GALK enzyme to establish a high-throughput biochemical assay for GALK activity in a 384-well microplate format at the HTS facility at the Broad Institute. As described in Materials and Methods, the 2-step luminescence-based high-throughput assay we developed measured GALK activity indirectly by quantifying the amount of ATP that remained in the GALK reaction. Therefore, the luminescent signal is inversely proportional to the amount of kinase activity.
During the assay development phase, we aimed to identify reaction conditions that would use the minimal amount of GALK enzyme, give a wide sensitivity range (i.e., high signal-to-noise ratio), and ensure that the reaction kinetics in the presence of a strong GALK inhibitor is within linear range. Using ATP ranging from 5 to 20 μM, we were unable to produce a low end-of-reaction value at 10 μM or higher of ATP, thus compromising dynamic separation. In fact, we established that the largest signal-to-noise ratio achieved among all tested conditions was 96-fold, which took place with 0.15 μg GALK protein in the presence of 5 μM ATP over a time course of 60 min at room temperature (). We therefore believe that this signal-to-noise ratio indicates good dynamic range, and we conclude that this assay provided the optimal conditions to identify strong inhibitors of GALK.
FIG. 3 Galactokinase (GALK) high-throughput screening (HTS) optimization. (a) Effect of varying amounts of adenosine triphosphate (ATP) and enzyme on GALK activity. GALK reaction was carried out at different ATP concentrations (5, 10, 20 μM) and different (more ...)
To test the robustness of the HTS assay, we performed Z′ factor analysis of the results. As illustrated in , a Z′ factor of 0.91 was determined from our assay. Therefore, we are confident that we have established a very robust, miniaturized, 2-step in vitro HTS assay for recombinant human GALK. As a proof of concept, we showed that an ATP mimetic, ATP-γ-S, acts as a GALK inhibitor by demonstrating decreased GALK activity in the presence of an increasing amount of ATP-γ-S ().
It is noteworthy that the 2-step luminescence-based assay executed in this case is unique because it measures the rate of disappearance of a substrate, rather than the formation of the product(s). Thus, the more luminescence detected at the endpoint, the less GALK activity was inferred. In addition, inhibitors that compete with ATP for luciferase, and therefore result in false negatives, will not be identified by this assay. Although these false negatives may be an undesirable outcome, one must realize that the compounds we are going to “miss out” in these cases are likely to be the less selective inhibitors that compete for ATP binding sites in other kinases. As a result, this assay may reduce, in the long run, the number of nonselective hits for GALK.
HTS of small-molecule inhibitors of GALK
Because of the mission of the HTS facility at the Broad Institute, it should be emphasized that the libraries of compounds we screened were not necessarily “drug-like.” Moreover, most compounds in the libraries were stored at a concentration of 10 mM. Consequently, the final concentration of the compounds in the assay was in the micromolar range. A Z′ factor assay was included each day as part of the quality control process, with a Z′ factor of 0.9 or above observed throughout all screening days. From the standard curves constructed with varying amount of GALK, we found that a composite z
-score of 7.68 was equivalent to 86.3% inhibition of GALK activity (data not shown). Hits were defined arbitrarily as a composite z
-score above 7.68 in combination with a reproducibility score of at least 0.7 (). Based on these criteria, 200 compounds, representing 0.4% of the total 50,000 compounds in the libraries, were found to inhibit GALK activity by 86.3% or more. This percentage is in good agreement with published results of primary HTS.32-34
FIG. 4 Selection of 150 compounds from primary screen. A composite z-score and a reproducibility score were assigned to each of the 50,000 compounds tested based on their ability to inhibit galactokinase (GALK) activity in vitro and the reproducibility of the (more ...)
In total, 150 of the 200 compounds were subsequently selected for repeated testing (i.e., confirmatory screen) under identical experimental conditions. Most hits not selected for confirmatory screen were compounds with a molecular weight in excess of 500. As a validation of our screen, we also identified a few known kinase inhibitors among our hits, whereas most of the others are novel compounds that have not been characterized.
shows the correlation results between the luminescence values recorded for each of the selected 150 compounds in the primary screen and the confirmatory screen. We were able to demonstrate a correlation coefficient of +0.48. Following completion of the confirmatory screen, 34 compounds were selected for further characterization (). The other 116 compounds were excluded either because inhibitory properties on GALK activity were not confirmed in the confirmatory screen or because of known toxicity.
FIG. 5 Primary versus confirmatory screen. Average (n = 2) luminescence values recorded for the top 150 compounds selected from the primary screen were plotted against average (n = 2) luminescence values of the corresponding compounds recorded at the confirmatory (more ...)
Structures of 34 selected compounds. Compounds are numbered by composite z-score in descending order.
Preliminary characterization of the selected compounds
The molecular weights of the 34 identified structures were between approximately 200 and 800. They cover a relatively wide range of lipophilicity based on calculated log octanol-water partition coefficients in the −1 to 8 range (CLOGP from ChemDraw 10, CambridgeSoft). Most compounds had no Lipinski violations, and 10 compounds had 1 violation, with a maximum of 2 violations in 3 compounds.
of the compounds studied thus far using the established luminescence-based assay ranged from 200 nM to 33 μM (data not shown). The IC50
values of 4 selected representatives (compounds 1, 3, 9, and 24) have also been confirmed in the standard pyruvate kinase/lactate dehydrogenase-coupled assay (). In this assay, ATP consumed by the galactokinase reaction is recycled by the pyruvate kinase reaction, and the amount of galactose present in the reaction is in large excess; hence, calculated reaction rates are close to initial velocity. Values obtained were somewhat lower than in the luminescence assay, which was in agreement with Wu and coworkers’ observation that IC50
s determined at conditions other than zero substrate conversion will be higher than the true rate.29
FIG. 7 Determination of IC50 for selected inhibitors. Dose response of the inhibition of galactokinase (GALK) activity in the standard pyruvate kinase/lactate dehydrogenase-coupled assay for compounds 1, 3, 9, and 24. Data are shown as a percentage of the corresponding (more ...)
All 34 compounds contain at least 2 rings, at least 1 of which is aromatic, and all have at least 2 hydrogen-bond acceptor sites. A pair of aromatic rings separated by 1, 2, or 3 bonds is the most common recurring motif. If we also consider rings flattened by at least 2 double bonds (and some resonance effects) such as those in structures 3, 5, 7, 8, 11, 13, 14, 16, 17, 21, 26, 27, 28, 31, 33, and 34, as a part of these pairs, then such a pair of close, flat rings is present in the majority of structures (31 of 34; separation of 1 bond: 1, 3, 5, 6, 8, 12, 13, 14, 15, 16, 19, 20, 21, 23, 24, 31, 34; 2 bonds: 3, 5, 6, 7, 9, 10, 17, 18, 19, 20, 21, 26, 28, 31, 34; 3 bonds: 1, 2, 5, 11, 12, 14, 16, 22, 23, 26, 27, 29, 30, 33). The remaining structures (4, 25, and 32) as well as some of the other ones (e.g., 1, 2, 6, 7, 23, 24, 30, 33) all contain an aromatic ring joined to a flat ring containing at least 1 >C =O moiety. One of the most frequent motifs is a 6-membered dihydro-thiazinone ring, usually thio and cyano substituted (e.g., 5, 11, 13, 16, or 27; 14 having a similar ring), as well as a 5-membered methylene-thiazolidinone ring (e.g., 3, 17, 26, 28). A cyano substituent is present in 7 of the 34 structures. Most structures are neutral, nonionizable compounds: only 1 contains an ionizable aliphatic amine (2), but 6 contain an aromatic carboxylic acid (10—isomeric, 12, 16, 22, 24, and 31). There are also quite a few halo-substituted phenols such as 9, 10, 18, and 30. Because there is considerable structural similarity among these 34 most promising compounds identified here, they can help in identifying an active binding site and serve as good leads for the development of an effective drug therapy for CG.
Our report represents the starting point of a series of ongoing studies that aimed to identify bioactive, potent, and selective GALK inhibitors. Our goal is that such inhibitors eventually will result in novel, medical treatment complementary to dietary therapy for patients with CG.